Self-adaption image binaryzation method based on residual image histogram cyclic shift

Abstract

The invention discloses a self-adaption image binaryzation method based on a residual image histogram cyclic shift. The method comprises the following steps that a target image is processed, and a corresponding background image is obtained; by analyzing the gray distribution characteristic of the target image, a corresponding binaryzation residual mask is obtained; by means of the background image and the target image, an initial residual image is obtained; a histogram of the initial residual image is counted, and a residual histogram is obtained; the horizontal axis of the histogram is the residual value, the vertical axis of the histogram is the number of pixels; an image in the residual value histogram conducts cyclic shift in the positive direction of the horizontal axis / residual value, an average residual value of the residual histogram subjected to cyclic shift is calculated to serve as a threshold value for conducting binaryzation operation on the residual image; the threshold value is used for conducting binaryzation on the residual image, and a binary image is obtained.

Description

technical field [0001] The invention relates to an image binarization method, in particular to an adaptive image binarization method based on the cyclic shift of the residual value image histogram. Relating to patent classification G06 Computing; Calculations; Counting G06K Data identification; Data representation; Record carriers; / 36 Image preprocessing, that is, image information processing without judging the identity of the image. G06K9 / 38 Quantization of analog image signals. Background technique [0002] In terms of target detection based on infrared sea surface images, image binarization technology has become the most common technical means in this field because of its huge advantages such as simple operation, high reliability, and good real-time performance. Generally speaking, image binarization technology includes two key factors: scope and binarization threshold. In practical applications, the selection of scope and the calculation of binarization threshold wil...

AI Technical Summary

Problems solved by technology

[0004] 1. The algorithm is susceptible to changes in the average gray level of the image

Due to factors such as different atmospheric radiation and different scene contents, it is difficult to guarantee uniform or similar gray distribution characteristics in the actual captured images. However, the existing algorithms only rely on the gray characteristics of the image for threshold segmentation. When the gray distribution characteristics of the image When changes occur, it is difficult to guarantee good detection results for images taken in different environments;

[0005] 2. The algorithm ignores the utilization of pixel local contrast information

Existing algorithms generally only use the absolute gray level of each pixel for segmentation, while ignoring the reference to the local contrast information of the pixel, which can easily lead to weak foreground pixels being missed when strong foreground pixels and weak foreground pixels appear in the image at the same time. inspection phenomenon;

[0006] 3. It is difficult to automatically calculate the optimal segmentation threshold

The existing threshold automatic calculation method can usually only get an ideal segmentation threshold when the foreground pixels have strong features (such as large area, strong gray scale, etc.), and when the foreground is small or its own gray scale is weak, The calculation result is difficult to guarantee a better segmentation effect

[0007] The existence of the above problems makes it difficult for existing image binarization methods to maintain high robustness in practical engineering applications.

Images